THE SILICON SHRINK

The computer will see you now.

Science journalist Oberhaus opens with a tragedy that illuminated a dark corner of medicine: in this case, the suicide of his sister, which prompted him to explore the use of PAI—psychiatric artificial intelligence—in mental health interventions and treatment. This “revolutionary New Thing,” as he calls it, has distinct advantages, at least in theory, over human practitioners: It can absorb huge quantities of information, discern patterns of behavior across populations, and perhaps detect when a psychic alarm bell is about to sound. On the other hand, as Oberhaus notes, there are 227 ways to be diagnosed with depression, while there are “more than 600,000 possible symptom combinations that could yield a diagnosis of post-traumatic stress disorder.” These are not the hard-and-fast results of scientific discovery, but instead the product of consensus built on “the ability to map these symptoms to internal dysfunction.” Given that AI, so far, has not been a reliable interpreter of emotional states or indeed of the meaning of phrases such as a teenager’s saying “I’m going to kill myself” after failing an exam, which may be serious but may also be hyperbole, it has obvious shortcomings. Yet, as Oberhaus observes, given the huge number of people who are suffering from mental disorders, with 20% of Americans experiencing anxiety requiring treatment, PAI is increasingly employed, with all its “potential for harm.” Oberhaus examines the history of PAI, with manifestations such as the popular echoic program ELIZA and newer technologies such as Facebook’s suicide prevention AI. He doesn’t entirely dismiss the possibility of machine intelligence being put to good use in the future, but this extended cautionary tale suggests that there’s still much work to be done before hailing PAI as “a new miracle cure.”

An eye-opening exposé of how machines are replacing people in a sphere they probably shouldn’t be.